Author
![]() |
Sudduth, Kenneth - Ken |
![]() |
Fraisse, Clyde |
![]() |
DRUMMOND, SCOTT - UNIV OF MO |
![]() |
Kitchen, Newell |
Submitted to: Geospatial Information in Agriculture and Forestry International Conference
Publication Type: Proceedings Publication Acceptance Date: 6/2/1998 Publication Date: N/A Citation: N/A Interpretive Summary: Site-specific management, or precision farming, is a strategy in which cropping inputs such as fertilizers are applied at varying rates across a field in response to variations in crop needs. To understand what these crop needs are from point to point, it is necessary to understand the relationship between crop yield and both controllable (such as fertilizer nutrients) and uncontrollable (such as topography) factors. The effect of these factors on yield is complex and may change from point to point within a field. We found that complex statistical techniques were able to satisfactorily model the relationship of yield to these other factors. We also found that an artificial neural network, a type of computer program that mimics the function of the human brain in a limited sense, was able to model yield. In addition, we evaluated the ability of an existing crop growth computer model, CROPGRO-Soybean, to model yield variations. We found that this model did not reproduce yields well because it did not account for the movement of water within the field by surface runoff. Ways of compensating for this shortcoming through the use of topographic analysis methods were developed but need more work for best performance. The results from this work will benefit scientists by providing them with new tools for investigating crop response to limiting factors. Producers and agribusiness will also benefit through improved fertilizer recommendations and management strategies developed with these techniques. Technical Abstract: Data collected on a 36-ha field in central Missouri were used to investigate methods for relating spatial grain yields to differences in those factors that can affect yields. Nonlinear, non-parametric data analysis methods, including projection pursuit regression and neural network analysis, were able to model yields as a function of soil and topographic data. Estimated yield maps obtained with these methods reproduced major patterns seen in actual yields. The use of cross validation techniques to guard against overfitting was important in obtaining reliable yield estimates. However, results obtained with these methods were not able to predict future years' yields and optimum management strategies, due to the uncertainties associated with year to year variations in climatic conditions. In order to more robustly evaluate yield-limiting factors across a range of climatic conditions, the CROPGRO Soybean model was used. The crop growth model was somewhat successful in representing spatial yield patterns but was deficient in not being able to account for yield variations due to excess water "run-on" from upland areas of the field. In an attempt to overcome this limitation, methods were developed to account for water redistribution based on soil and topographic characteristics. |